High-fidelity adaptive curvelet domain primary-multiple separation processing of seismic data

摘要

Methods and systems for separating multiple events from primary events in noisy seismic data are described. Multiples are predicted and then the predictions are improved by least-square matching filtering in the space and time domain. An adaptive curvelet domain separation (ACDS) is then performed and the ACDS equation is solved with an iterative soft-thresholding technique. Further processing can be added to compensate for prediction inaccuracy or variable/excessive seismic data noise by dividing the seismic data into predetermined bands and processing each band independently.